The development of efficient spatial methods and the widespread availability of spatially referenced data have changed the landscape of exposure assessment in disease etiology. Most spatial-based exposure-disease assessment approaches require knowing where study participants are located in space and time, and linking that information to spatially-referenced data that estimate potential for exposure. Such approaches not only allow for estimation of both current and past exposures, but are efficient alternatives to traditional methods of collecting exposure data longitudinally, enhancing the utility of existing large cohorts by reverse engineering exposures when improved exposure surfaces become available. However, while the spatial resolution of exposure surfaces has greatly improved, our ability to locate people in space (with geocoding) has not, and remains a rate- limiting factor in accurate exposure assessment. The effort engaged in improving spatially referenced exposure data is compromised without addressing the problem of misclassification of the location of people (geocoding uncertainty). We have developed a geocoding approach that records the exact spatial extent of the final geocode that fully describes the area in which the study participant is known to be located, and a novel statistical approach to incorporate variability in exposure and covariate data based on spatial extent. These combined approaches can be extended to appropriately incorporate spatial uncertainty from geocoding misclassification into the overall exposure assessment model. We aim to test the inclusion of geocode uncertainty into an exposure assessment model, and then apply that approach in a study of the role of pesticides in childhood leukemia ? an example which has a high resolution exposure surface, high variability in geocode accuracy, and is an excellent example of some of the worst case scenarios in assuming uniform spatial certainty of geocodes. We will ensure wide dissemination of the approach as a global solution to the problem of misclassified study participant geolocation.